Towards a farm-scale digital twin for carbon farming: site-specific implementation, validation, and potential for future projections

Increasing soil organic carbon (SOC) stocks in agricultural soils can enhance soil health, support higher crop productivity with fewer inputs, and contribute to offsetting greenhouse gas emissions. SOC can be increased through carbon-farming practices, such as cover crops, residue retention, and carbon-rich organic amendments. However, SOC changes are difficult to monitor at relevant temporal and spatial scales, and existing modelling and data-driven approaches often struggle to capture the complex interactions between management, soil, and future climate conditions.

In this poster, we provide details on the implementation of a farm-scale digital twin, defined as an accurate and dynamic representation of a real-world agricultural system that is consistently updated, evaluates future scenarios, and provides actionable insights for stakeholders. The current version of the digital twin supports decision-making for carbon farming and agricultural management while accounting for key interactions within the soil-crop-atmosphere continuum. It is built on the process-based agroecosystem model AgroC, which couples SoilCO₂ for water, heat, gas, and solute transport in a one-dimensional soil column, SUCROS for organ-specific crop growth, and RothC for soil carbon turnover. The digital twin was applied to 14 fields of varying size (1.5–15 ha) from a farm in western Germany. Field-specific crop rotation, seeding and harvest dates, fertilization, and application of mushroom compost were implemented in close collaboration with the farmer, enabling simulation of SOC dynamics over periods of 5 to 16 years depending on field data availability.

Although within-field soil heterogeneity can be represented using spatially distributed simulations, each field was modelled using a single soil unit, as regional soil maps indicated relatively homogeneous conditions. This assumption was supported by Electromagnetic Induction (EMI) measurements performed on two representative fields in August 2025. For model validation, SOC data from previous years were available for eight fields. Additionally, ten fields were sampled between August and November 2025 at 56 locations, resulting in 164 soil samples for which SOC was estimated via the loss-on-ignition method. Further validation was performed using harvest data. The digital twin simulations well matched measured values, and it was confirmed that the sustainable practices implemented by the farmer had positively influenced SOC trajectories over the study period. Other unique information could be provided to the farmer, such as the rate at which SOC would decrease if regenerative practices were interrupted or the relative importance of individual actions in each field, with residue retention being the most prominent.

Taken together, these results show that the developed farm-scale digital twin allows to account for complex interactions within the soil–crop system, can provide holistic, tailored analyses with the potential to not only support SOC management, but also adapt agricultural practices to climate change, improve water regulation, and enhance soil ecosystem functions for sustainable agriculture.